import bnlearn as bn
bif_file= 'sprinkler'
bif_file= 'alarm'
bif_file= 'andes'
bif_file= 'asia'
bif_file= 'pathfinder'
bif_file= 'sachs'
bif_file= 'miserables'
bif_file= 'filepath/to/model.bif'# Load asia DAG
model = bn.import_DAG('asia')
# plot ground truth
G = bn.plot(model)
# Sampling
df = bn.sampling(model, n=10000)
# Structure learning of sampled dataset
model_sl = bn.structure_learning.fit(df, methodtype='hc', scoretype='bic')
# Plot based on structure learning of sampled data
bn.plot(model_sl, pos=G['pos'])
# Compare networks and make plot
bn.compare_networks(model, model_sl, pos=G['pos'])[bnlearn] >Import <asia>
[bnlearn] >Downloading example [asia] dataset..
[bnlearn] >Extracting files..
[bnlearn] >Loading bif file </opt/anaconda3/lib/python3.8/site-packages/bnlearn/data/asia.bif>
[bnlearn] >Checking CPDs..
[bnlearn] >Check for DAG structure. Correct: True
[bnlearn] >Plot based on BayesianModel
png
[bnlearn] >Forward sampling for 10000 samples..
0%| | 0/8 [00:00<?, ?it/s]
[bnlearn] >Computing best DAG using [hc]
[bnlearn] >Set scoring type at [bic]
[bnlearn] >Plot based on BayesianModel
[bnlearn] >Existing coordinates from <pos> are used.
png
[bnlearn] >Existing coordinates from <pos> are used.
png
png
(array([[51, 5],
[ 4, 4]]),
target asia tub smoke lung bronc either xray dysp
source
asia 0 -1 0 0 0 0 0 0
tub 2 0 0 0 0 1 0 0
smoke 0 0 0 -1 -1 0 0 0
lung 0 0 2 0 0 1 0 0
bronc 0 0 2 0 0 0 0 -1
either 0 0 0 0 2 0 1 1
xray 0 0 0 0 0 0 0 0
dysp 0 0 0 0 2 0 0 0)import csv
import codecs
from itertools import islicedf| asia | tub | smoke | lung | bronc | either | xray | dysp | |
|---|---|---|---|---|---|---|---|---|
| 0 | 1 | 1 | 0 | 1 | 1 | 1 | 1 | 0 |
| 1 | 1 | 1 | 0 | 1 | 1 | 1 | 1 | 1 |
| 2 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 |
| 3 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 |
| 4 | 1 | 1 | 0 | 1 | 1 | 1 | 1 | 1 |
| … | … | … | … | … | … | … | … | … |
| 9995 | 1 | 1 | 0 | 1 | 1 | 1 | 1 | 1 |
| 9996 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 |
| 9997 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 |
| 9998 | 1 | 1 | 0 | 1 | 1 | 1 | 1 | 1 |
| 9999 | 1 | 1 | 0 | 1 | 1 | 1 | 1 | 1 |
10000 rows × 8 columns
df.to_csv('synthetic_asia.csv')# Load sprinkler DAG
model = bn.import_DAG('sprinkler')
# plot ground truth
G = bn.plot(model)
# Sampling
df = bn.sampling(model, n=10000)
# Structure learning of sampled dataset
model_sl = bn.structure_learning.fit(df, methodtype='hc', scoretype='bic')
# Plot based on structure learning of sampled data
bn.plot(model_sl, pos=G['pos'])
# Compare networks and make plot
bn.compare_networks(model, model_sl, pos=G['pos'])[bnlearn] >Import <sprinkler>
[bnlearn] >Checking CPDs..
[bnlearn] >Check for DAG structure. Correct: True
[bnlearn] >Plot based on BayesianModel
png
[bnlearn] >Forward sampling for 10000 samples..
0%| | 0/4 [00:00<?, ?it/s]
[bnlearn] >Computing best DAG using [hc]
[bnlearn] >Set scoring type at [bic]
[bnlearn] >Plot based on BayesianModel
[bnlearn] >Existing coordinates from <pos> are used.
png
[bnlearn] >Existing coordinates from <pos> are used.
png
png
(array([[8, 4],
[2, 2]]),
target Cloudy Sprinkler Rain Wet_Grass
source
Cloudy 0 1 1 2
Sprinkler 0 0 0 -1
Rain 0 2 0 -1
Wet_Grass 0 2 2 0)df.to_csv('synthetic_sprinkler.csv')# Load earthquake DAG
model = bn.import_DAG('alarm')
# plot ground truth
G = bn.plot(model)
# Sampling
df = bn.sampling(model, n=10000)
# Structure learning of sampled dataset
model_sl = bn.structure_learning.fit(df, methodtype='hc', scoretype='bic')
# Plot based on structure learning of sampled data
bn.plot(model_sl, pos=G['pos'])
# Compare networks and make plot
bn.compare_networks(model, model_sl, pos=G['pos'])[bnlearn] >Import <alarm>
[bnlearn] >Downloading example [alarm] dataset..
[bnlearn] >Extracting files..
[bnlearn] >Loading bif file </opt/anaconda3/lib/python3.8/site-packages/bnlearn/data/alarm.bif>
[bnlearn] >Checking CPDs..
[bnlearn] >Warning: CPD [BP] does not add up to 1 but is: [[1. 1. 1.]
[1. 1. 1.]
[1. 1. 1.]]
[bnlearn] >Warning: CPD [HREKG] does not add up to 1 but is: [[0.9999999 0.9999999 1. ]
[0.9999999 1. 1. ]]
[bnlearn] >Warning: CPD [HRSAT] does not add up to 1 but is: [[0.9999999 0.9999999 1. ]
[0.9999999 1. 1. ]]
[bnlearn] >Warning: CPD [PRESS] does not add up to 1 but is: [[[1. 1. 1. 1.]
[1. 1. 1. 1.]]
[[1. 1. 1. 1.]
[1. 1. 1. 1.]]
[[1. 1. 1. 1.]
[1. 1. 1. 1.]]]
[bnlearn] >Check for DAG structure. Correct: True
[bnlearn] >Plot based on BayesianModel
png
[bnlearn] >Forward sampling for 10000 samples..
0%| | 0/37 [00:00<?, ?it/s]
[bnlearn] >Computing best DAG using [hc]
[bnlearn] >Set scoring type at [bic]
[bnlearn] >Plot based on BayesianModel
[bnlearn] >Existing coordinates from <pos> are used.
png
[bnlearn] >Existing coordinates from <pos> are used.
png
png
(array([[1304, 19],
[ 15, 31]]),
target LVFAILURE HISTORY LVEDVOLUME CVP PCWP HYPOVOLEMIA \
source
LVFAILURE 0 1 -1 0 0 0
HISTORY 0 0 0 0 0 0
LVEDVOLUME 2 0 0 -1 1 2
CVP 0 0 2 0 0 0
PCWP 0 0 0 0 0 0
HYPOVOLEMIA 0 0 -1 0 0 0
STROKEVOLUME 0 0 0 0 0 2
ERRLOWOUTPUT 0 0 0 0 0 0
HRBP 0 0 0 0 0 0
HR 0 0 0 0 0 0
ERRCAUTER 0 0 0 0 0 0
HREKG 0 0 0 0 0 0
HRSAT 0 0 0 0 0 0
ANAPHYLAXIS 0 0 0 0 0 0
TPR 0 0 0 0 0 0
ARTCO2 0 0 0 0 0 0
EXPCO2 0 0 0 0 0 0
VENTLUNG 0 0 0 0 0 0
INTUBATION 0 0 0 0 0 0
MINVOL 0 0 0 0 0 0
FIO2 0 0 0 0 0 0
PVSAT 0 0 0 0 0 0
VENTALV 0 0 0 0 0 0
SAO2 0 0 0 0 0 0
SHUNT 0 0 0 0 0 0
PULMEMBOLUS 0 0 0 0 0 0
PAP 0 0 0 0 0 0
PRESS 0 0 0 0 0 0
KINKEDTUBE 0 0 0 0 0 0
VENTTUBE 0 0 0 0 0 0
MINVOLSET 0 0 0 0 0 0
VENTMACH 0 0 0 0 0 0
DISCONNECT 0 0 0 0 0 0
CATECHOL 0 0 0 0 0 0
INSUFFANESTH 0 0 0 0 0 0
CO 0 0 0 0 0 0
BP 0 0 0 0 0 0
target STROKEVOLUME ERRLOWOUTPUT HRBP HR ... PRESS KINKEDTUBE \
source ...
LVFAILURE 1 0 0 0 ... 0 0
HISTORY 0 0 0 0 ... 0 0
LVEDVOLUME 2 0 0 0 ... 0 0
CVP 0 0 0 0 ... 0 0
PCWP 0 0 0 0 ... 0 0
HYPOVOLEMIA -1 0 0 0 ... 0 0
STROKEVOLUME 0 0 0 0 ... 0 0
ERRLOWOUTPUT 0 0 1 0 ... 0 0
HRBP 0 0 0 0 ... 0 0
HR 0 0 1 0 ... 0 0
ERRCAUTER 0 0 0 0 ... 0 0
HREKG 0 0 0 0 ... 0 0
HRSAT 0 0 0 0 ... 0 0
ANAPHYLAXIS 0 0 0 0 ... 0 0
TPR 0 0 0 0 ... 0 0
ARTCO2 0 0 0 0 ... 0 0
EXPCO2 0 0 0 0 ... 0 0
VENTLUNG 0 0 0 0 ... 0 2
INTUBATION 0 0 0 0 ... 1 0
MINVOL 0 0 0 0 ... 0 0
FIO2 0 0 0 0 ... 0 0
PVSAT 0 0 0 0 ... 0 0
VENTALV 0 0 0 0 ... 0 0
SAO2 0 0 0 0 ... 0 0
SHUNT 0 0 0 0 ... 0 0
PULMEMBOLUS 0 0 0 0 ... 0 0
PAP 0 0 0 0 ... 0 0
PRESS 0 0 0 0 ... 0 0
KINKEDTUBE 0 0 0 0 ... 1 0
VENTTUBE 0 0 0 0 ... 1 2
MINVOLSET 0 0 0 0 ... 0 0
VENTMACH 0 0 0 0 ... 0 0
DISCONNECT 0 0 0 0 ... 0 0
CATECHOL 0 0 0 1 ... 0 0
INSUFFANESTH 0 0 0 0 ... 0 0
CO 0 0 0 0 ... 0 0
BP 0 0 0 0 ... 0 0
target VENTTUBE MINVOLSET VENTMACH DISCONNECT CATECHOL \
source
LVFAILURE 0 0 0 0 0
HISTORY 0 0 0 0 0
LVEDVOLUME 0 0 0 0 0
CVP 0 0 0 0 0
PCWP 0 0 0 0 0
HYPOVOLEMIA 0 0 0 0 0
STROKEVOLUME 0 0 0 0 0
ERRLOWOUTPUT 0 0 0 0 0
HRBP 0 0 0 0 0
HR 0 0 0 0 0
ERRCAUTER 0 0 0 0 0
HREKG 0 0 0 0 0
HRSAT 0 0 0 0 0
ANAPHYLAXIS 0 0 0 0 0
TPR 0 0 0 0 1
ARTCO2 0 0 0 0 1
EXPCO2 0 0 0 0 0
VENTLUNG 2 0 0 0 0
INTUBATION 0 0 0 0 0
MINVOL 2 0 0 0 0
FIO2 0 0 0 0 0
PVSAT 0 0 0 0 0
VENTALV 0 0 0 0 0
SAO2 0 0 0 0 -1
SHUNT 0 0 0 0 0
PULMEMBOLUS 0 0 0 0 0
PAP 0 0 0 0 0
PRESS 0 0 0 0 0
KINKEDTUBE 0 0 0 0 0
VENTTUBE 0 0 2 2 0
MINVOLSET 0 0 -1 0 0
VENTMACH -1 2 0 0 0
DISCONNECT -1 0 2 0 0
CATECHOL 0 0 0 0 0
INSUFFANESTH 0 0 0 0 -1
CO 0 0 0 0 0
BP 0 0 0 0 0
target INSUFFANESTH CO BP
source
LVFAILURE 0 0 0
HISTORY 0 0 0
LVEDVOLUME 0 0 0
CVP 0 0 0
PCWP 0 0 0
HYPOVOLEMIA 0 0 0
STROKEVOLUME 0 1 0
ERRLOWOUTPUT 0 0 0
HRBP 0 0 0
HR 0 1 0
ERRCAUTER 0 0 0
HREKG 0 0 0
HRSAT 0 0 0
ANAPHYLAXIS 0 0 0
TPR 0 0 1
ARTCO2 0 0 0
EXPCO2 0 0 0
VENTLUNG 0 0 0
INTUBATION 0 0 0
MINVOL 0 0 0
FIO2 0 0 0
PVSAT 0 0 0
VENTALV 0 0 0
SAO2 0 0 0
SHUNT 0 0 0
PULMEMBOLUS 0 0 0
PAP 0 0 0
PRESS 0 0 0
KINKEDTUBE 0 0 0
VENTTUBE 0 0 0
MINVOLSET 0 0 0
VENTMACH 0 0 0
DISCONNECT 0 0 0
CATECHOL 0 0 0
INSUFFANESTH 0 0 0
CO 0 0 1
BP 0 0 0
[37 rows x 37 columns])df.to_csv('synthetic_alarm.csv')# Load sachs DAG
model = bn.import_DAG('sachs')
# plot ground truth
G = bn.plot(model)
# Sampling
df = bn.sampling(model, n=10000)
# Structure learning of sampled dataset
model_sl = bn.structure_learning.fit(df, methodtype='hc', scoretype='bic')
# Plot based on structure learning of sampled data
bn.plot(model_sl, pos=G['pos'])
# Compare networks and make plot
bn.compare_networks(model, model_sl, pos=G['pos'])[bnlearn] >Import <sachs>
[bnlearn] >Downloading example [sachs] dataset..
[bnlearn] >Extracting files..
[bnlearn] >Loading bif file </opt/anaconda3/lib/python3.8/site-packages/bnlearn/data/sachs.bif>
[bnlearn] >Checking CPDs..
[bnlearn] >Warning: CPD [Akt] does not add up to 1 but is: [[1. 1. 1. ]
[1. 1. 0.99999998]
[0.99999992 1. 1. ]]
[bnlearn] >Warning: CPD [Erk] does not add up to 1 but is: [[1. 1. 1. ]
[1. 1. 1. ]
[1. 1. 0.99999999]]
[bnlearn] >Warning: CPD [Jnk] does not add up to 1 but is: [[1. 1. 1. ]
[1.00000001 0.99999993 1. ]
[0.99999997 1. 1. ]]
[bnlearn] >Warning: CPD [Mek] does not add up to 1 but is: [[[1. 1. 1. ]
[1. 1. 1. ]
[1. 1. 0.9999999 ]]
[[1.00000002 1. 1. ]
[0.99999994 1. 1. ]
[1.00000001 1. 0.9999999 ]]
[[0.99999996 1. 1. ]
[1. 1. 1. ]
[1. 1. 0.9999999 ]]]
[bnlearn] >Warning: CPD [P38] does not add up to 1 but is: [[1. 1. 1. ]
[1. 1.00000003 1. ]
[0.99999999 1. 1. ]]
[bnlearn] >Warning: CPD [PIP2] does not add up to 1 but is: [[0.99999997 1. 1. ]
[0.99999999 1. 1. ]
[1. 1.00000001 1.00000001]]
[bnlearn] >Warning: CPD [PIP3] does not add up to 1 but is: [1.0000001 1. 1.0000001]
[bnlearn] >Warning: CPD [PKA] does not add up to 1 but is: [0.9999999 1.00000001 1. ]
[bnlearn] >Warning: CPD [PKC] does not add up to 1 but is: 0.9999999999999999
[bnlearn] >Warning: CPD [Raf] does not add up to 1 but is: [[1. 1. 1. ]
[1. 1. 1.00000001]
[1.00000001 1. 1. ]]
[bnlearn] >Check for DAG structure. Correct: True
[bnlearn] >Plot based on BayesianModel
png
[bnlearn] >Forward sampling for 10000 samples..
0%| | 0/11 [00:00<?, ?it/s]
[bnlearn] >Warning: Computing DAG with 11 nodes can take a very long time!
[bnlearn] >Computing best DAG using [hc]
[bnlearn] >Set scoring type at [bic]
[bnlearn] >Plot based on BayesianModel
[bnlearn] >Existing coordinates from <pos> are used.
png
[bnlearn] >Existing coordinates from <pos> are used.
png
png
(array([[97, 7],
[ 7, 10]]),
target Erk Akt PKA Mek Jnk PKC Raf P38 PIP3 PIP2 Plcg
source
Erk 0 1 0 0 0 0 0 0 0 0 0
Akt 0 0 0 0 0 0 0 0 0 0 0
PKA 1 1 0 -1 1 2 -1 1 0 0 0
Mek 1 0 2 0 0 2 0 0 0 0 0
Jnk 0 0 0 0 0 0 0 0 0 0 0
PKC 0 0 -1 -1 1 0 -1 1 0 0 0
Raf 0 0 2 1 0 2 0 0 0 0 0
P38 0 0 0 0 0 0 0 0 0 0 0
PIP3 0 0 0 0 0 0 0 0 0 1 2
PIP2 0 0 0 0 0 0 0 0 0 0 2
Plcg 0 0 0 0 0 0 0 0 -1 -1 0)df.to_csv('synthetic_sachs.csv')